Alexa, show me a good voice UI
summary
1 in 6 American homes have a voice device, so design decisions matter: favor narrow domains and repetitive, action-oriented flows (ordering pizza, checking weather, requesting an Uber), choose between voice-only, voice-first, and voice-added platforms, expect speech-to-text ~95% accurate, use explicit/implicit confirmations based on risk, apply progressive disclosure and decaying context, avoid ambiguous yes/no prompts, and leverage dialogic pairs and insertion sequences; Alexa Prize demos hit ~12 minutes average conversation but general skills work best when focused and test-driven.
AI-written summary · about
chapters (30)
- 0:00Intro
- 0:18Should I be building for voice
- 2:50Voice platforms
- 5:28Leave the user
- 5:40Yes or no questions
- 6:13User training
- 7:20Grices Maxim
- 7:35Quantity
- 8:04Too much information
- 8:38Quality
- 8:53Lying
- 9:16Relation
- 9:56Google Home
- 10:23Manor
- 10:35Voice Partner
- 11:07Speak out loud
- 11:31Ellipsis
- 12:02Context
- 12:30Implicit
- 12:54Decaying Context
- 13:27Confirmation
- 13:53Explicit
- 14:18Inferential
- 14:32Trust
- 15:02Progressive disclosure
- 15:32Power users
- 16:02Correction
- 16:45Dialogic Pairs
- 17:12Results
- 17:38Next
related talks
44mDelivering On The Promise Of Voice With Alexa
13mVOICE Summit 2019 | VED Talks | Dylan Zwick | July 23
48mTen Things Every Voice Application Should Do
52mThe most natural UI. Building voice-enabled experiences with Amazon Alexa
24mVoice design best practices
16mWeak signals of a vocal revolution in UIs
38mConversational UIs: Talking to Siri, Alexa, and Your Web Browser
44mDeveloping Skills & Actions for Voice Assistants